I've seen a lot of demos, and I've never seen high quality materials generated by these products. Teachers who've used LLMs to "level" readings have told me the results were unusable. Colleagues have given up on using these products because it's a slot machine of garbage. Who's checking for quality?
The government is quietly testing AI tools in prisons to generate profile reports on offenders. Decisions impacting people's lives, rehabilitation, and release require human oversight, strict transparency, and absolute accountabilityânot algorithmic trial and error.
www.thestar.com/politics/fed...
This creates a perverse incentive: humanizing AI boosts user engagement and shields developers from accountability when things go wrong.
Shout out to my amazing advisors on this project, @minzlicht.bsky.social and Jason Plaks.
Link to paper:
online.ucpress.edu/collabra/art...
I'm very excited to hear what the authors of this paper have to say about critical resistance to the AI project, but was annoyed to see that ACM offers an "AI podcast" of the paper. I'm guessing the authors didn't agree to this.
How is a teacher creating a personal piece of ed tech for a student "computer science"? I can't believe that we're now being asked to use ed tech to create more ed tech. Maybe the CS concept covered is recursion? đ€·ââïž
Part of an effective resistance has to be holding the line on how we use language. Every time we say "AI" in reference to technological applications, we lose ground.
Alternatives:
Products marketed as "AI"
Software-generated text
Chatbots
Statistical models
Classifiers
Data-based programming
LLMs
"Critical thinking: AI use is to build, not replace, student thinking." What makes the Toronto School Board believe this is feasible, or even possible?
This type of statement is meaningless in practice and only serves to push the assumption that these products are educationally valuable.
Real world "AI literacy" = product adoption, as always.
Reduced cognitive load comes from fluency/practice. Context switching to mete out tasks to software *is* cognitive load (see calculator use). Also, you're not "deep thinking" if you're not doing the thinking! :o
The long term cognitive effects may not be known yet, but it sounds like Toronto is well on its way to finding out!
ElizabethwithaZ
K-8:
- Adaptive reading platforms to adjust to reading level
- Text to speech/voice tools - long established, now AI "enhanced"
- Introduction to AI concepts (Google's Teachable Machine course)
- Teachers drafting coursework or rubrics etc (not visible to students)
